the Prepare recipe is almost entirely a row-by-row affair, so it doesn't blend well with computing things like a median, which take many rows into consideration. You can
- run the Prepare recipe using Spark (or SQL if the input dataset is SQL)
- use a notebook or a python recipe
Using only visual recipes in the flow, you can also retrieve the median with a window recipe (to compute the quantiles) but it requires some convolutions afterwards to get the median value back on the rows.